Can not run pipeline: Can't identify image sets

Usage & Issues
#1

Hi, our lab is new to CellProfiler and we are trying to learn how to use the software for our needs. I have been trying all day to analyze one image and test the software out but I haven’t been able to get off the ground because it keeps telling me that it “hasn’t been able to identify image sets”… I have my image loaded in the file list and an analysis module loaded to run. Can someone tell me what I’m missing? I have attached an image of my screen. Notice how it starts to processing my image in the lower right hand corner though… but it never completes, both sides of the time only increase as time continues.

Thanks so much!

Kelly
Scripps Florida

Screen Shot 2015-03-12 at 4.43.59 PM.png1706×958 126 KB

#2

Hi Kelly,

Could you post the project file associated with this screenshot, so we can take a look?

Thanks!
-Mark

#3

Hi Mark,

By the project file, do you mean the pipeline or the image itself?

Here is the image below:

The pipeline’s i’ve tried to use were both from the website download for “example percent positive” and one pipeline that i found here on the forums. I was just trying to start with one single image for positive stained cell count. Our images are in color - do I need to convert them to gray scale before I place them in Cell Profiler?

Thank you!

Kelly
Scripps Florida

#4

Hi Kelly,
Yes, the pipeline would help us (File -> Export -> Pipeline). An no, you don;t need to convert the images from color to grayscale (or vice-versa) outside of CellProfiler

I would guess that you should look at the NamesAndTypes module and ensure that “Select the image type” is set to “Color”. But if that doesn’t immediately help, then post your pipeline here.

Cheers,
David

#5

Hi Mark,

The pipelines I’ve tried are below.

We are looking at images that are stained with DAPI and are positively stained for two or three other stains (FITC, 594, and 647). Our goal is to count total DAPI cells and from that population determine how many are positive for one, two, or all three stains. I’ve been looking through the forums to find pipelines or explanations about how other people have gone about these calculations, which seem do-able, but I keep running into the issue where cell profiler won’t be able to identify my images… even when I add them to the images > file list and names and types > update. What I thought was causing the issue was that I’m not associating the images correctly since I may have all 3 or 4 channels in one image? Do you need to separate each channel into individual images and convert them to grey scale? I tried this and I wasn’t successful moving past names and types even though I was just trying to correlate each channel to a separate image. It would be ideal if we could analyze our 40x or 100x images with all channels in one image.

Is this possible? What information would I not be putting into the input module about my images that is causing the analysis to not move forward?

Thanks!

Kelly
DLpipeline.cppipe (10.3 KB)
DLpipeline(1).cppipe (10.3 KB)
ExamplePercentPositive.cppipe (15.7 KB)

#6

And this pipeline below as well since we are interested in stain intensity differences between cells. I would also like to analyze a full image without cropping as well, if possible. I’m not sure how cell profiler determines where to crop and analyze an image.

Thank you!

Kelly
PipelineF_DLogan.cppipe (16.5 KB)

#7

Hi Kelly,

To answer your original question: Yes, the reason you aren’t able to pick up any image sets is that NamesAndTypes is not configured correctly. This module needs to mach the filename in order to assign the image a name; since the rules specified for the various pipelines you tried do not match your specific files, the pipeline fails to start.

I see that the image you uploaded has the nuclei in gray but another stain in color. I’m not sure what you did in terms of merging the individual channels, but unfortunately it renders the image unusable. Ideally, you want the color image to represent onnly onne stain for each color. But since the nuclei are gray, they appear in all 3 and hence cannot be separated (or at least, cannot be deconvolved from the other stains).

You are better served by saving each of the channels (DAPI, FITC, 594, and 647) as a individual grayscale file, and giving each channel a specific nomenclature such that NamesAndTypes can specify rules to match the filename text and assign it to a name. You can see examples of how this is done on our examples page: cellprofiler.org/examples.shtml, where most of the examples there contain multi-channel assays.

If you can provide individual channels for a site of interest, we would be in a better position to produce a pipeline that can do what you need.

Regards,
-Mark

#8

I’m a newbie poster, so first of all I want to say that I am greatly impressed with CellProfiler and appreciate the effort that has gone into getting it this far. May it live long and prosper!

I’m resurrecting this thread because I am seeing the message quoted in the subject line. I created and tested a new pipeline to identify cell outlines from one channel image in a LSM file and to use the outlines to quantify fluorescence intensity in another channel image in the same LSM file. It all worked very nicely.

In preparation of passing the pipeline on to someone who is actually going to use it with lots of their own LSM files, I exported the pipeline to a .cppipe file. Then I did a final test by double clicking on the .cppipe file to start CellProfiler with the pipeline loaded, but without the LSM file that was part of the original project. After dragging the test LSM file into the file drop area, I pressed ‘Analyze Images’ and got the message: “Cannot run pipeline. … The pipeline did not identify any image sets.”.

However, on the basis of checking previous messages here, I tried pressing ‘Update metadata’ in the Metadata module and then it worked as expected.

I would prefer this updating happened automatically and I don’t have to tell the user they need to do this every time they load new data. Is this correct behavior for the program? Is there anything I can do to make it work as I would like?

Thanks,
Francis

#9

I’m not honestly sure if that’s expected behavior- I also would’ve assumed that the metadata would update automatically. Unfortunately loading metadata from the image headers is a time consuming and occasionally buggy process (between supporting all the manufacturer’s various formats and the fact that the metadata is inherently more corruptible than the image from what I understand about a lot of file formats, there are a lot of places for things to go wrong!) so
a) I think metadata extraction isn’t done unless specifically asked for and
b) It’s usually a good idea to check and make sure that all of your files have been correctly loaded, matched, etc before proceeding onto running the pipeline.

If it’s enough of a problem for your user, you/they can request a change to the behavior here, but hopefully it’s a minimal annoyance.

Glad you otherwise enjoy the program!

#10

Thanks for the feedback and suggestions.

It would be up to me to request a change in behaviour. As it is just now, the user is very happy that they no longer have to analyse their data manually in ImageJ, which is what they were previously doing. So my guess is that having to do two more mouse clicks is an entirely acceptable price to pay for a greatly streamlined workflow.

That said, I might request a way to force metadata updating when the pipeline is Run. If doing so might cause other problems, it could be made an option.

In any case, it might be worth changing the error message to suggest making sure the metadata information is updated (by pressing “Update metadata” in the Metadata module/page). As it is currently, it refers to a “configuration problem” - something a naive user might be confused by.

Francis